Cleanup

rknarukulla · Jan 8, 2014 · 775b7b1 · 775b7b1
1 parent 9c05ffe
commit 775b7b1
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 2 deletions.
diff --git a/experiments/src/main/java/samples/NeuralNetMnistPretrain.java b/experiments/src/main/java/samples/NeuralNetMnistPretrain.java
@@ -33,7 +33,6 @@ public static void main(String[] args) throws Exception {
       ls[i].momentum_stable = .99f;
       //ls[i].l1 = .005f;
       ls[i].init(ls, i);
-      if (i>=1) ls[i].randomize(new java.util.Random(), 1.0f);
     }
     return ls;
   }

diff --git a/src/main/java/hex/Layer.java b/src/main/java/hex/Layer.java
@@ -96,7 +96,7 @@ public final void init(Layer[] ls, int index) {
     init(ls, index, true, 0, new MersenneTwisterRNG(MersenneTwisterRNG.SEEDS));
   }
 
-  public void init(Layer[] ls, int index, boolean weights, long step, Random rand) {
+  protected void init(Layer[] ls, int index, boolean weights, long step, Random rand) {
     _a = new float[units];
     _e = new float[units];
     _previous = ls[index - 1];
@@ -121,6 +121,8 @@ public void init(Layer[] ls, int index, boolean weights, long step, Random rand)
    */
   // cf. http://machinelearning.wustl.edu/mlpapers/paper_files/AISTATS2010_GlorotB10.pdf
   public void randomize(Random rng, float prefactor) {
+    if (_w == null) return;
+
     if (initial_weight_distribution == InitialWeightDistribution.UniformAdaptive) {
       final float range = prefactor * (float)Math.sqrt(6. / (_previous.units + units));
       for( int i = 0; i < _w.length; i++ )

diff --git a/src/main/java/hex/NeuralNet.java b/src/main/java/hex/NeuralNet.java
@@ -150,6 +150,9 @@ void startTrain() {
         case Tanh:
           ls[i + 1] = new Layer.Tanh(hidden[i]);
           break;
+        case TanhWithDropout:
+          ls[i + 1] = new Layer.TanhDropout(hidden[i]);
+          break;
         case Rectifier:
           ls[i + 1] = new Layer.Rectifier(hidden[i]);
           break;
@@ -169,6 +172,7 @@ void startTrain() {
       ls[i + 1].momentum_stable = (float) momentum_stable;
       ls[i + 1].l1 = (float) l1;
       ls[i + 1].l2 = (float) l2;
+      ls[i + 1].max_w2 = max_w2;
       ls[i + 1].loss = loss;
     }
     if( classification )